Classification of ECG signals using LDA with factor analysis method as feature reduction technique.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Cardiology

Background:

• Electrocardiogram (ECG) signal analysis, particularly the QRS complex, is crucial for diagnosing cardiac dysfunctions.
• Wavelet coefficients derived from the QRS complex serve as effective features for ECG analysis.
• Feature reduction techniques are essential for improving the efficiency and accuracy of cardiac arrhythmia classification.

Discussion:

• This study investigates the efficacy of various factor analysis rotation methods (varimax, equimax, quartimax) for reducing ECG QRS complex features.
• The Principal Component Method is employed to estimate component loadings during the feature reduction process.
• Linear Discriminant Analysis (LDA) is utilized as the classifier for distinguishing between different heartbeats.

Key Insights:

• The equimax rotation method, applied to wavelet coefficients of the QRS complex, achieved the highest average accuracy of 99.056% on unknown datasets.
• Feature reduction using factor analysis significantly enhances the classification performance of cardiac arrhythmias.
• Comparative analysis of accuracy, sensitivity, and positive predictivity validates the superiority of the equimax rotation technique.

Outlook:

• Further research could explore advanced feature extraction and reduction methods for more complex cardiac conditions.
• Integration of these techniques into real-time ECG monitoring systems holds potential for early disease detection.
• Validation on larger and more diverse patient populations is recommended to confirm generalizability.